Lore

Geographic experimentation

Geographic experimentation randomizes by geography to measure incrementality. Spend is increased in selected regions while other regions are held constant as controls. Outcomes are compared, and the control regions serve as the counterfactual.

How to apply

It is a specific implementation of the Test vs. control incrementality experiment and answers the Incrementality and the counterfactual problem. Two ways to build the control: matched-market-testing and synthetic-controls. A worked workflow is in geolift-test-workflow.

Attribution and incrementality converge at small scale and diverge as the brand grows. A dashboard that was trustworthy early can silently become misleading (see Deterministic (in-platform) attribution).